The impact that market fluctuations have on the demand and the willingness to pay in freight transport

Abstract

[ES] La gestión de ingresos (RM) se refiere a la teoría y la práctica de la gestión de la demanda con apoyo de las tecnologías de la información por medios como los precios o la disponibilidad de los productos basados en modelos de demanda para maximizar los beneficios o los ingresos. Evaluar y garantizar una estrategia de gestión de los ingresos puede ayudar a las empresas a comprender el comportamiento de los clientes y obtener información útil para maximizar sus beneficios. Sin embargo, a menudo esto supone un reto para empresas con grandes bases de datos..

Por ello, nos proponemos los modelos o herramientas basadas en aprendizaje automático mas adecuadas contenedores. Los productos básicos de bajo valor son más sensibles a las subidas de precios, ya que los costes de transporte representan una parte bastante importante del precio de la mercancía. El desarrollo de un modelo que aproveche esta información y encuentre correlaciones entre los datos puede ayudar a Maersk a imponer precios recomendados por sus servicios, maximizando sus ingresos como objetivo final

[EN] Background: implementing an optimized revenue management strategy has been determined for companies when it comes to maximizing their profits. Maersk has been struggling to achieve this goal since their strategy is based on assumptions from their expertise in the industry instead of facts. They believe that there could be a connection between market price and demand for some commodities. Data and Methods: this study uses information provided by Maersk containing all type of information about bookings and corresponding commodities. The purpose is to develop the most suitable Data­Driven model in order to identify how the value of the commodity shipped impacts the customers’ willingness to pay for the container and sensitivity towards changes in container prices. In addition, TreeExplainer is implemented to interpret and assess which features impact more in the model prediction at a local and global level. Three different applications with distinct commodities and fixing different origins and destinations are carried out so as to investigate how features behave when the underlying commodity change. Results: XGBoost is chosen as the most suitable machine learning model to apply in this problem. After assessing different models, it presents the highest accuracy­interpretability trade­off. On the other hand, macro­trends and geographical parameters appear to have a significant impact on customers’ willingness to pay for a container and, accordingly, the pricing strategy. Container prices are found to change depending on the commodity shipped and they are negatively correlated with the demand. Conclusion: this investigation has confirmed that Maersk’s intuition was true, discovering a correlation between market prices and demand. The developed model captures causation pretty accurately and learns a lot from what happened in the past to develop a better understanding of what could occur in the future for similar situation. Therefore, t brings Maersk helpful information in its quest to establish the most optimized pricing strategy in order to maximize revenue.